Relative paths by default.

[theemathas]

I believe that absolute paths by default in use statements are a bad idea.

[pnkfelix]

Notes from previous discussion of this topic at a team meeting : https://github.com/rust-lang/meeting-minutes/blob/master/weekly-meetings/2014-02-25.md#relative-ref--req-leading--in-use

(and here is the ticket that that discussion refers to; this predates the RFC repo so it was a ticket on the rust repo tagged with the RFC label. rust-lang/rust#10910 )

[glaebhoerl]

Rendered

[theemathas]

@pnkfelix I didn't see that discussion before. However, the resolution of that discussion seems unclear to me.

Additionally, I think that doing it this way would help users to better understand modules, since you just have to do everything the same way as you do in a typical file system, but with :: instead of /

[nikomatsakis]

Regardless of the merits of the proposal, I think it is simply too late to make a breaking change of this magnitude.

[theemathas]

@nikomatsakis I believe that my proposal does not break as much code as the int/uint -> isize/usize and deriving -> derive changes.

Changes in user code would be able to be mostly (if not entirely) automated. Even if there is a lot of non-automate-able changes, you can just add a feature gate for the old behavior.

Also note that there aren't any backwards compatibility promises for 1.0 alpha.
(From http://blog.rust-lang.org/2014/12/12/1.0-Timeline.html)

A pre-release version indicates that the version is unstable and might not
satisfy the intended compatibility requirements as denoted by its associated
normal version.

Actually, I don't know of any valid reasons not to make this change. (Except if the compiler requires a lot of modification, and I believe that it doesn't)

Please correct me if I am wrong.

[sinistersnare]

valid reasons not to make this change

• nobody wants to implement it
• people want absolute paths in imports
• too much craziness going on in the compiler to do it in time for 1.0.

There's three.

[sinistersnare]

-1, changing semantics like this so close to 1.0 might not be a good idea.

[yedpodtrzitko]

would this mean that modules called super or crate would collide on import?

[theemathas]

@yedpodtrzitko IIRC modules named super or crate are already compile errors.

[theemathas]

@nikomatsakis @sinistersnare I have realized that I misinterpreted the 1.0 alpha compatibility guarantees. I am retracting this RFC.

[bombless]

Relative-path including reminds me of PHP file including, which is a pain in the neck for me.